Some time i am getting this message in console.

My kernel:

So what i need to do, and how can i change this setting?
I never got this messege with the kernel 2.6.8-2-386

I am using EM64T P-IV (3.0 GHz with 2 MB l2 cache) in debian

Could be a few things.
Are the fans working on the cpus?
If so are they big enough for the cpu capacity?
Is there plenty of air entering and leaving the box?
Are any vent opens blocked?
Is it in a cabinet with door closed? If so leave door open.
Can be an underpowered power supply or a powersupply going bad.

Now what tempertures are you getting when this happens?

I would say it gets this info from lmsensors package. You might need to edit /etc/sensors.conf to match your hardware. Run the command ' sensors ' to see what it is seeing.

Lets go from here.
Brian1

I am having this problem as well. I can easily recreate it on many of my servers by running (among other more true to life ways).
sensors is installed, and if I kill (CTRL+C) dd when the message pops up and run sensors, I get output similar to the following:
[HTML]it87-isa-0290
Adapter: ISA adapter
fan1: 6490 RPM (min = 3013 RPM, div = 8)
fan2: 4440 RPM (min = 3013 RPM, div = 8)
CPU Temp: +61°C (low = +15°C, high = +70°C) sensor = diode
[/HTML]
The high is set to 70 and the warning appears to be going off at 60. Furthermore, the high was previously set to 65 and the end result was no different (the process didn't run longer or anything). Everything I have read (which isn't that much) says to adjust sensors for the processor, but from the looks of my trials, I don't think sensors is what is actually telling the kernel the temperature threshold has been reached... Output of sensors prior to running the command above:
[HTML]it87-isa-0290
Adapter: ISA adapter
fan1: 6490 RPM (min = 3013 RPM, div = 8)
fan2: 4440 RPM (min = 3013 RPM, div = 8)
CPU Temp: +38°C (low = +15°C, high = +65°C) sensor = diode
[/HTML]
I am running sensors as quickly as I can by pressing (CTRL+C)(CTRL+SHIFT+V) in a terminal window as fast I can as soon as I see the message pop up. I have "sensors(ENTER)" in the clipboard and "sensors" shows up in the terminal before the next prompt and executes immediately.
I tried one more time with the threshold set to 75 and the same thing still happened; the temperature immediately afterwards was 61.
The server I used for testing had the same issues with large scripts. Reseating the heatsink with fresh thermal paste made no difference. This server is running CentOS 2.6.9-34.ELsmp but was running a much older version when the problem originally started showing up. I am having similar problems on several of my servers and would love to know how to convince the kernel to let the CPU run warmer than 60C°. TIA

are the fans even kicking on? if not, acpi may not be built into the kernel

My servers are 1U rack servers, all fans are always on full speed. Turns out the p4thermal interrupt kicks in at 60°C. I'm recompiling the kernel without p4thermal interrupt support (/usr/src/kernels/2.6.9-34.EL-smp-i686/include/config/x86/mce/p4thermal.h) right now, hopefully that resolves the problem. I will try to remember to post an update when I'm done. =)

Great,
I am using a EM64T 3.0 GHz processor. How can i see the heat of the processor.

It was a pain, as CentOS 4.3 uses a 2.6.9 kernel whose source has some misformatted QLogic source files, but I got it done, and now I don't get those messages or the associated slowdown. The CPU doesn't appear to get much hotter under full load either. Can anyone verify for me whether or not lm_sensors will actually tell the kernel to throttle the CPU down in case of overheating?

lavluda, try running sensors-detect as root. If it doesn't run, you need to get lm_sensors and install it. If it does, then you can go through it and follow the on screen instructions. Afterwards, running sensors should tell you all of the information that lm_sensors is can see. That data won't necessarily be accurate, and you may have to modify sensors.conf (followed by running sensors -s) a few times, but if it is accurate, or if you get it straightened out by tinkering, then you'll be good to go. If there is no info, on the other hand, all you can do is wait for updates and see if drivers for your chips have been added to lm_sensors. Good luck.